Sentiment Analysis Flashcards
7 cards from real NLP practice questions. Tap to flip, then mark Knew It or Still Learning — missed cards come back until you master them.
Read the first 7 Sentiment Analysis flashcards as text
What role does 'transfer learning' play in modern sentiment analysis?
Answer: It leverages pretrained language model weights to improve sentiment classification with less labeled data
Transfer learning fine-tunes pretrained models like BERT on labeled sentiment data, achieving high accuracy with fewer examples.
Which sentiment analysis technique uses a graph of words to propagate sentiment scores?
Answer: Graph-based sentiment propagation
Graph-based propagation spreads known sentiment scores from seed words through a word co-occurrence or similarity graph to label unlabeled words.
What is 'emotion detection' and how does it differ from polarity classification?
Answer: Emotion detection identifies specific emotions like joy or anger, while polarity only classifies positive or negative
Polarity classification assigns positive/negative/neutral labels, while emotion detection maps text to discrete emotion categories such as Ekman's six basic emotions.
What is the 'Pointwise Mutual Information' (PMI) method used for in sentiment analysis?
Answer: Estimating semantic orientation of phrases by comparing co-occurrence with positive vs. negative seed words
PMI measures how much more often a phrase co-occurs with positive seed words versus negative ones to estimate its sentiment orientation.
Which data augmentation technique is commonly applied to improve sentiment classifier robustness?
Answer: Synonym replacement, random insertion, or back-translation to create varied training examples
Techniques like synonym replacement and back-translation generate new training sentences that preserve sentiment while varying wording.
In attention-based models for sentiment analysis, what does the attention mechanism highlight?
Answer: Words most relevant to determining the sentiment of the sentence
Attention weights indicate which tokens the model focuses on most when predicting sentiment, often highlighting opinion words and negations.
What challenge does 'cross-lingual sentiment analysis' address?
Answer: Building sentiment models that work across languages, especially for low-resource languages without labeled data
Cross-lingual sentiment analysis aims to transfer sentiment knowledge from high-resource languages (like English) to languages with little or no labeled sentiment data.